Re: [bug-gawk] Computed regex and getline bug / issue

[arnold]

"Andrew J. Schorr" <address@hidden> wrote:

> Hi Arnold,
>
> On Sat, May 10, 2014 at 10:50:52PM +0300, Aharon Robbins wrote:
> > Indeed. Here is the fix for that too.
> > 
> > Got the other test case, I'll push it.
>
> Great.  Thanks.  But with this new patch, is the previous one
> actually required?  I tried backing it out, as per the attached patch,
> and it passes "make check" for me.
>
> Regards,
> Andy

Hi Andy. The previous change may not be required but it will buy us
some pico-efficiencies since we won't end up trying the read again a
few times and hitting EOF each time we pull more stuff out of the
buffer until we actually hit the end of the buffer. In the worst case
of say a 4096 byte buffer, with a zero character record and one character
separator, we could save 4000+ read calls that tell us we've hit EOF.

That's a worst case scenario and is likely to be not nearly so bad
in practice.  :-)

The code is a little clearer as it stands now; I think I'd like to
leave it.

But that was an interesting question to think about.

Thanks!

Arnold